Research Of3PCF Parallel Algorithm Based On MPI

3-Point Correlation Function is one of the most important cosmology simulationalgorithms, how to solve it more efficiently is one of the problems that puzzledastronomers. In this paper we proposed a parallel algorithm and several optimizedmethods based on FFT’s algorithm, in order to improve its precision and calculatingability.In order to promote the precision, the algorithm divides large-scale matrix whichcan get the high accuracy results into small matrix, which can be easily computed inthe range of the capacity of memory.In order to promote the efficiency, the paper proposed the distributed model. Themain node assigns the computing tasks to each node evenly. Each node computes thetasks independently. At last, the main node will combine all the sub results. Thecomputing ability will scale up with the computing nodes increasing. From the results,we can find that the speed up is linear with processes increasing.At last, the paper verifies the validity of parallel algorithm. With further study inthe algorithm, we design several optimized experiments: First, filtering the zeromatrix; Second, removing the unnecessary zeroes in filter matrix; Third,3D FFTmethod changing into1D FFT method; Fourth, the length of every dimension can bedivided into2,3,5,7. The results show that the methods improve the efficiency of3-Point Correlation Function.The experimental results show that the expansion of compute nodes andoptimized methods improve the computational efficiency of the parallel algorithms,blocking large-scale matrix can achieve high accuracy results of the calculation.